Designing Experiments for Large Numbers of Factors

In many practical problems, for example in industrial and scientific research, the number of factors that might possibly influence the response of interest can very large. Sometimes scientific understanding, or results from previous experiments, can be used to reduce the list of possible factors for investigation to a size that allows economic experimentation in terms of cost and time. Where such prior knowledge is not available, a variety of other approaches can be used to identify the important factors. And hence screen out unimportant factors, through a series of experiment. This paper briefly describes various screening techniques, and discusses their advantages and disadvantages with particular emphasis on detecting important interactions in industrial product development. Recent work on investigating group screening methods will be described, and criteria discussed for assessing and choosing between designs.